Re: [OSM-talk] [Tagging] - RFC -Motorway_linkimpliesoneway=??
In your letter dated Wed, 08 Oct 2008 12:25:53 -0400 you wrote: >Philip Homburg <[EMAIL PROTECTED]> writes: >> But seriously, what I expect is a set of rules of the form "if in-germany >> then highway=trunk implies oneway=yes". > >This leads to a nightmare. Those rules would need to be implemented >in every tool that works with OSM data (and cares about oneway >properties). And? All code also has to have an XML parser. I think it can be just a library that gets as input a way (or a node) and returns a set of default tags. >I am for the simple aproach: If it is oneway - tag it as oneway. I've no problem with that. Just that the absence of the oneway is taken to be 'oneway=no'. Anyhow, I think that three types application applications that need to know about defaults: - rendering engines, - consistency checkers. - routing programs For rendering engines it is not all that important if the defaults are not exactly right (we don't really need any arrows in dual-carriage roads, most people know they are oneway anyhow). Consistency checkers are complicated anyhow. And may also have a lot of heuristics. >If we really must have implied values for certain properties, keep >that simple as well and don't make in dependant on too many tags >(like: it's oneway if it is in Germany, but not in Saxony or on an >island and only if the speed limit is above 80 km/h, or, if it is in >France and ...) [I know I am exaggerating.] And the third class, the routing programs, really needs information about speed limits. And that is going to require some kind of localization. And once you got the mechanism in place, you can just as well use it for all kinds of access restrictions. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC -Motorway_linkimpliesoneway=??
2008/10/9 Matthias Julius <[EMAIL PROTECTED]>: > > This leads to a nightmare. Those rules would need to be implemented > in every tool that works with OSM data (and cares about oneway > properties). > It's a nightmare we're probably going to have to address at some point if we want to do good routing. There are many traffic rules that are local, affect routing, and are not explicitly signed on the roads. Here's an example. In Queensland (and I think the rest of Australia) it is illegal to do a U-turn at traffic lights unless there is a sign that specifically says you can. It is legal to U-turn at road junctions without traffic lights unless there is a sign that says you can't. If you want a routing application to take this into account, we're going to either have a local rule that says "Don't U-turn at lights etc" or get every intersection with lights mapped with U-turn restrictions in every direction that aren't signed in the real world (good luck with that). Stephen ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC -Motorway_linkimpliesoneway=??
Philip Homburg <[EMAIL PROTECTED]> writes: > In your letter dated Wed, 8 Oct 2008 10:07:40 +0200 you wrote: >>On Wed, Oct 8, 2008 at 9:29 AM, Philip Homburg >><[EMAIL PROTECTED]> wrote: >>> [...] >>> I don't think that a relation should be used to imply oneway=yes. It's just >>> too risky. >>> In a country where just all trunk roads are dual-carriage ways, defaulting >>> to oneway=no is just too risky. >> >>having different defaults for different countries is going to be a >>problem anyway, since not every developer is going to know every >>country default > > Most developers don't speak Italian either :-) > > But seriously, what I expect is a set of rules of the form "if in-germany > then highway=trunk implies oneway=yes". This leads to a nightmare. Those rules would need to be implemented in every tool that works with OSM data (and cares about oneway properties). I am for the simple aproach: If it is oneway - tag it as oneway. If we really must have implied values for certain properties, keep that simple as well and don't make in dependant on too many tags (like: it's oneway if it is in Germany, but not in Saxony or on an island and only if the speed limit is above 80 km/h, or, if it is in France and ...) [I know I am exaggerating.] And implied values should be clearly indicated on Map Features not only on the individual tag pages. Matthias ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
2008/10/8 Philip Homburg <[EMAIL PROTECTED]>: > I don't think this is about local interpretations. It is about having > safe defaults. Agreed 100% > Of course, adding oneway=no to all trunk ways that do not have a oneway tag > can be done by a script. Clearly it could be - but it certainly _shouldn't_ be, unless that script can be confined to an area where it is certain that all trunks really are dual in accordance with local tagging norms. But I prefer this... > Another approach, that may also work for trunk roads is to write a consistency > checker that tries to detect this situation. > > A first pass tries to find roads with the same names, or unnamed roads that > are roughly parallel. Then report any road in that set that doesn't have an > explicit oneway tag. This is a lot safer IMHO, since (arguably) if we have mapped only one carriageway of a dualled road, it's more useful for routing software to be allowed to route both ways over it. With incomplete mapping, insisting on correct setting of oneway isn't all that useful. > I don't think that a relation should be used to imply oneway=yes. It's just > too risky. A dual-carriageway relation should. In a world where not even roundabouts are guaranteed to be one-way, we can at least trust a dual-carriageway to be so. > In a country where just all trunk roads are dual-carriage ways, defaulting > to oneway=no is just too risky. > > But you seem to care more about the burden of retagging some existing trunk > roads than about having safe defaults. Not at all. I don't care about the burden of retagging trunk roads. But I don't want the tail to wag the dog. The trunk tag was conceived for roads that are not inherently dual-carriageway. Established practice is to explicitly tag the special (and recognisable) case of dual-carriageways with oneway=yes. What I'm saying is that dualled section of trunk highways that are not yet explicitly tagged oneway should now be so tagged. Alternatively, we can introduce highway=gelbe_autobahn that implies oneway and bulk retag the German trunks. > On the other hand, given that localization is likely to happen eventually > anyhow, it may at some point become just a local decision. If we could reliably determine the nationality of a section of road I'd be a lot more relaxed about this matter (and others, like maxspeed). And once that day arrives, as it surely will, we can happily zap tags deemed on a regional level to be implicit. But we're not there yet. Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
In your letter dated Wed, 8 Oct 2008 10:07:40 +0200 you wrote: >On Wed, Oct 8, 2008 at 9:29 AM, Philip Homburg ><[EMAIL PROTECTED]> wrote: >> [...] >> I don't think that a relation should be used to imply oneway=yes. It's just >> too risky. >> In a country where just all trunk roads are dual-carriage ways, defaulting >> to oneway=no is just too risky. > >having different defaults for different countries is going to be a >problem anyway, since not every developer is going to know every >country default Most developers don't speak Italian either :-) But seriously, what I expect is a set of rules of the form "if in-germany then highway=trunk implies oneway=yes". There are two issues: encoding the rules and classifying the roads. Roads can, for example, be classified by having areas with highwaycode=D tags, optionally with bots that copy highwaycode tags down to ways with a highway tag to make the whole system more robust. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
On Wed, Oct 8, 2008 at 9:29 AM, Philip Homburg <[EMAIL PROTECTED]> wrote: > [...] > I don't think that a relation should be used to imply oneway=yes. It's just > too risky. > In a country where just all trunk roads are dual-carriage ways, defaulting > to oneway=no is just too risky. having different defaults for different countries is going to be a problem anyway, since not every developer is going to know every country default IMHO the safer way is to tag dual-carriage ways in some explicit way, either with oneway=yes or with some other tag that _always_ implies oneway (like the junction=roundabout one): the relation may be one such solution In Italy we are adding the explicit oneway tag even to motorways, just to be on the safe side -- Elena of Valhalla homepage: http://www.trueelena.org email: [EMAIL PROTECTED] ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
In your letter dated Tue, 7 Oct 2008 14:34:48 +0100 you wrote: >2008/10/7 Philip Homburg <[EMAIL PROTECTED]>: >> I'm a bit worried about routing software sending people the wrong way up >> a dual-carriage way. I very much prefer to default to a safe state. And that >> means either requiring explicit yes/no oneway tags for both motorway and >> trunk or implying oneway for those roads. > >Only the explicit tagging is a valid choice, then. I'm certainly not >retagging the bulk of my national road network to oneway=no just >because of shifts in local interpretations of what a trunk road is. I don't think this is about local interpretations. It is about having safe defaults. Of course, adding oneway=no to all trunk ways that do not have a oneway tag can be done by a script. >And it still doesn't solve the problem of dual-carriageway primary, >secondary or tertiary roads, of which there are plenty. That's true. Another approach, that may also work for trunk roads is to write a consistency checker that tries to detect this situation. A first pass tries to find roads with the same names, or unnamed roads that are roughly parallel. Then report any road in that set that doesn't have an explicit oneway tag. >> I think this is risky: if one way or another the dual_carriageway relation >> is not there, then routing software will default to an unsafe configuration. > >I think you're misunderstanding me here - nothing will protect us from >broken tagging. But I'm saying that the mapper should have a choice. >Either explicitly tag your carriageways as oneway or, if you value >tidiness, provide a relation. It's my counter-offer to those who say >"all trunks in my country are dual, therefore I want to reduce clutter >by implying oneway". I'm saying "No, if you want to reduce clutter, do >so by using the relation. This way, we solve the problem of clutter >for all road types, but we don't invalidate existing trunks.". I don't think that a relation should be used to imply oneway=yes. It's just too risky. In a country where just all trunk roads are dual-carriage ways, defaulting to oneway=no is just too risky. But you seem to care more about the burden of retagging some existing trunk roads than about having safe defaults. On the other hand, given that localization is likely to happen eventually anyhow, it may at some point become just a local decision. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
2008/10/7 Philip Homburg <[EMAIL PROTECTED]>: > I think local norms are fine. However that requires a lot of localization > work. > > But a global norm is better than a local one. > > Localization is likely to happen anyway when people start displaying speed > limits. Or do you want to tag even the smallest country road with the > appropriate speed limit for all types of vehicles? > > Maybe technical solutions are an option: defining administrative areas that > contain the defaults that apply. This works in theory, but we are some way off it in practice. You either need to tell every segment of road which administrative area it lies within (can be difficult) _or_ to get your boundaries dead right (this is proving very tricky with today's sources). Most end-devices will probably be too stupid to apply such rules on the fly, though pre-processing is certainly an option. > I'm a bit worried about routing software sending people the wrong way up > a dual-carriage way. I very much prefer to default to a safe state. And that > means either requiring explicit yes/no oneway tags for both motorway and > trunk or implying oneway for those roads. Only the explicit tagging is a valid choice, then. I'm certainly not retagging the bulk of my national road network to oneway=no just because of shifts in local interpretations of what a trunk road is. And it still doesn't solve the problem of dual-carriageway primary, secondary or tertiary roads, of which there are plenty. >>those that are members of a dual_carriageway relation > > I think this is risky: if one way or another the dual_carriageway relation > is not there, then routing software will default to an unsafe configuration. I think you're misunderstanding me here - nothing will protect us from broken tagging. But I'm saying that the mapper should have a choice. Either explicitly tag your carriageways as oneway or, if you value tidiness, provide a relation. It's my counter-offer to those who say "all trunks in my country are dual, therefore I want to reduce clutter by implying oneway". I'm saying "No, if you want to reduce clutter, do so by using the relation. This way, we solve the problem of clutter for all road types, but we don't invalidate existing trunks.". Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
In your letter dated Tue, 7 Oct 2008 11:04:44 +0100 you wrote: >2008/10/7 Philip Homburg <[EMAIL PROTECTED]>: > >> For trunk roads, it might be just a safe default to assume that the way is >> oneway, unless tagged explicitly as single-carriage (oneway=no). > >People can keep saying that, but it won't make it true :) > >Trunk roads shouldn't be implied oneway, simply because of the >established usage of this tag to represent normal single-carriageway >roads. (National primary routes in Ireland, primary A roads in UK, >possibly others). My contention is that implicit tagging is only valid >where we have a global norm. Local norms aren't enough. I think local norms are fine. However that requires a lot of localization work. But a global norm is better than a local one. Localization is likely to happen anyway when people start displaying speed limits. Or do you want to tag even the smallest country road with the appropriate speed limit for all types of vehicles? Maybe technical solutions are an option: defining administrative areas that contain the defaults that apply. >Based on this reasoning, if you choose to accept it, it ISTM that only >motorway mainlines and roundabouts should be assumed oneway. > >BTW, there may be an alternative resolution to this matter. The >problem here is that we can't agree, country-to-country, which highway >tags should generally imply _dual-carriageway_. However, we do have >ways of recognising a dual-carriageway from other clues - either the >existence of parallel ways with the same ref and different directions >or (better) the existence of a dual_carriageway relation. I'm a bit worried about routing software sending people the wrong way up a dual-carriage way. I very much prefer to default to a safe state. And that means either requiring explicit yes/no oneway tags for both motorway and trunk or implying oneway for those roads. >those that are members of a dual_carriageway relation I think this is risky: if one way or another the dual_carriageway relation is not there, then routing software will default to an unsafe configuration. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
2008/10/7 Ed Loach <[EMAIL PROTECTED]>: > This is a roundabout where you can go either way around the "big" > roundabout, and at each junction with a road there is a > mini-roundabout. When these sort of roundabouts, often nicknamed > "Magic Roundabouts" > http://en.wikipedia.org/wiki/Magic_Roundabout_(Hemel_Hempstead) > cropped up on this other email list I wondered how they were tagged > in OSM, because of the implied oneway of junction=roundabout. http://www.openstreetmap.org/?lat=51.562812&lon=-1.77143&zoom=18&layers=B000FTF This is the original Magic Roundabout in Swindon. Here, some licence has been taken with the ways, as you'll see that the "two-way" nature of the flow is implemented here as separate one-way ways. On some sections, this is accurate, as there are real traffic islands. In others, there is just hatching, but since you may not cross it, it's not the biggest tagging sin imaginable. Strictly speaking, the mini-roundabouts should be tagged as such, but I can imagine that would be difficult to do nicely given the dual ways that would converge at each one. Road geeks would contend that Magic Roundabouts are not really roundabouts (apart from the outer minis), but "gyratory traffic systems". The distinction being that they diverge from standard roundabout rules of flow, right-of-way, layout etc., but are still circular. The API seems to be antisocial at the moment, so I can't check the tagging, but my inclination here would be not to tag anything but the 5 outer roundabouts with junction=roundabout. Think of what a routing application might make of it... Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
On Tue, Oct 7, 2008 at 12:04 PM, Dermot McNally <[EMAIL PROTECTED]> wrote: >[...] > So we can solve this problem right now by deciding that a better list > of implied-one-way highway ways is: > > those with junction=roundabout > those that are members of a dual_carriageway relation > > That will cover any stretches of trunk or motorway (or anything else) > that happen to be dualled. +1 to this solution: it seems the safer one and explicit tagging is not a big deal anyway -- Elena of Valhalla homepage: http://www.trueelena.org email: [EMAIL PROTECTED] ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
Dermot wrote: > So we can solve this problem right now by deciding that a > better list > of implied-one-way highway ways is: > > those with junction=roundabout Funnily enough we were discussing roundabouts on a non-OSM related list yesterday, and I looked up Greenstead Roundabout in Colchester http://www.openstreetmap.org/?lat=51.88448&lon=0.93284&zoom=17&layer s=0B00FFF This is a roundabout where you can go either way around the "big" roundabout, and at each junction with a road there is a mini-roundabout. When these sort of roundabouts, often nicknamed "Magic Roundabouts" http://en.wikipedia.org/wiki/Magic_Roundabout_(Hemel_Hempstead) cropped up on this other email list I wondered how they were tagged in OSM, because of the implied oneway of junction=roundabout. In the Colchester example it is a circular primary route with 5 mini-roundabouts at the junctions. I wonder whether junction=roundabout oneway=no would have had the same results? Or would renderers think to check for oneway=no on roundabouts? Ed ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
2008/10/7 Philip Homburg <[EMAIL PROTECTED]>: > For trunk roads, it might be just a safe default to assume that the way is > oneway, unless tagged explicitly as single-carriage (oneway=no). People can keep saying that, but it won't make it true :) For me, there are very few cases where oneway should safely be implied, and I generally tag these explicitly regardless: motorway (almost universally so, but we have the option to set oneway=no where not the case). motorway_link and other *_link (though we know that there are many countries where two-way stretches of these are common) anything with junction=roundabout Trunk roads shouldn't be implied oneway, simply because of the established usage of this tag to represent normal single-carriageway roads. (National primary routes in Ireland, primary A roads in UK, possibly others). My contention is that implicit tagging is only valid where we have a global norm. Local norms aren't enough. Based on this reasoning, if you choose to accept it, it ISTM that only motorway mainlines and roundabouts should be assumed oneway. BTW, there may be an alternative resolution to this matter. The problem here is that we can't agree, country-to-country, which highway tags should generally imply _dual-carriageway_. However, we do have ways of recognising a dual-carriageway from other clues - either the existence of parallel ways with the same ref and different directions or (better) the existence of a dual_carriageway relation. So we can solve this problem right now by deciding that a better list of implied-one-way highway ways is: those with junction=roundabout those that are members of a dual_carriageway relation That will cover any stretches of trunk or motorway (or anything else) that happen to be dualled. Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
>For trunk roads, it might be just a safe default to assume that the way is >oneway, unless tagged explicitly as single-carriage (oneway=no). Another safe option is for routing application to ignore all motorway, motorway_link, trunk, and trunk_link ways that are not explicitly tagged as either oneway=yes or oneway=no. Then the JOSM/Validator can show missing oneway tags as errors. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
In your letter dated Mon, 6 Oct 2008 19:04:28 +0100 you wrote: >2008/10/6 Matthias Julius <[EMAIL PROTECTED]>: >> "Stephen Hope" <[EMAIL PROTECTED]> writes: >> >>> Bad assumption. This may be the case in parts of Europe and the USA, >>> but certainly not in most parts of the world. >> >> Maybe not in most parts of the worlds, but most trunk roads. ;-) > >I know that the German mappers have decided that "trunk" is a handy >extra road category that can be used for Schnellstra=DFen. I also know >that other countries find it a useful classification for similar >purposes. But the country whose road system gave us the "trunk" tag >has thousands of km of standard single-carriageway road that bears the >tag. I would assume that oneway sections are in the minority in the >UK. In Ireland they certainly are. I wonder if it would be worth while setting up a wiki to collect the legal definitions of road types in the various countries. For example, the Dutch traffic regulations do not say anything about whether is motorway is dual-carriage or not. However they are explicitly one-way. So the Dutch equivalent of a trunk road is usually a single-carriage way but it is also oneway. However, in this case the oneway-ness is more or less irrelevant for OSM. It only means that you can't turn on the road, and that you can't drive in reverse. In practice, all Dutch motorways are dual-carriage. So it makes sense to assume oneway-ness by default. For trunk roads, it might be just a safe default to assume that the way is oneway, unless tagged explicitly as single-carriage (oneway=no). ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
Stefan Monnier <[EMAIL PROTECTED]> writes: >> There will still be people who miss all that and keep on tagging as >> before, but I don't see a reasonable way of avoiding that - except not >> changing implied values. > > You could change editors so as to automatically display the "default" > value of missing tags (as well as automatically remove tags that are set > to their default value). Do you really want your editor to display for every highway: oneway=no; bridge=no; tunnel=no; access=yes; area=no; lit=no; disused=no; wheelchair=no; boat=no; ... ? Matthias ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
2008/10/6 Matthias Julius <[EMAIL PROTECTED]>: > "Stephen Hope" <[EMAIL PROTECTED]> writes: > >> Bad assumption. This may be the case in parts of Europe and the USA, >> but certainly not in most parts of the world. > > Maybe not in most parts of the worlds, but most trunk roads. ;-) I know that the German mappers have decided that "trunk" is a handy extra road category that can be used for Schnellstraßen. I also know that other countries find it a useful classification for similar purposes. But the country whose road system gave us the "trunk" tag has thousands of km of standard single-carriageway road that bears the tag. I would assume that oneway sections are in the minority in the UK. In Ireland they certainly are. I can't argue with your conclusion, though - I always explicitly tag oneway where it applies. Usually even on roundabouts :P Dermot -- -- Iren sind menschlich ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] [Tagging] - RFC - Motorway_linkimpliesoneway=??
"Stephen Hope" <[EMAIL PROTECTED]> writes: > Bad assumption. This may be the case in parts of Europe and the USA, > but certainly not in most parts of the world. Maybe not in most parts of the worlds, but most trunk roads. ;-) I just wanted to show that if we start to base default values on the highest probability of the possible values we will end up with endless discussions and an endless confusion. Matthias ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk